Rapid image deconvolution and multiview fusion for optical microscopy

The contrast and resolution of images obtained with optical microscopes can be improved by deconvolution and computational fusion of multiple views of the same sample, but these methods are computationally expensive for large datasets. Here we describe theoretical and practical advances in algorithm...

Full description

Saved in:

Bibliographic Details
Published in	Nature biotechnology Vol. 38; no. 11; pp. 1337 - 1346
Main Authors	Guo, Min, Li, Yue, Su, Yijun, Lambert, Talley, Nogare, Damian Dalle, Moyle, Mark W., Duncan, Leighton H., Ikegami, Richard, Santella, Anthony, Rey-Suarez, Ivan, Green, Daniel, Beiriger, Anastasia, Chen, Jiji, Vishwasrao, Harshad, Ganesan, Sundar, Prince, Victoria, Waters, Jennifer C., Annunziata, Christina M., Hafner, Markus, Mohler, William A., Chitnis, Ajay B., Upadhyaya, Arpita, Usdin, Ted B., Bao, Zhirong, Colón-Ramos, Daniel, La Riviere, Patrick, Liu, Huafeng, Wu, Yicong, Shroff, Hari
Format	Journal Article
Language	English
Published	New York Nature Publishing Group US 01.11.2020 Nature Publishing Group
Subjects	631/1647/245/2225 631/1647/328 631/1647/328/2237 Acceleration Agriculture Algorithms Animals Bioinformatics Biomedical and Life Sciences Biomedical Engineering/Biotechnology Biomedicine Biotechnology Brain - diagnostic imaging Caenorhabditis elegans - embryology Cell Line Computer applications Control Convolutional codes Data fusion Datasets Deconvolution Deep Learning Embryos Humans Image contrast Image processing Image Processing, Computer-Assisted Innovations Life Sciences Light filters Light microscopy Light sheets Machine learning Mice Microscope and microscopy Microscopes Microscopy Optical microscopes Optical microscopy Point spread functions Spatial discrimination Spatial resolution Zebrafish - embryology United States
Online Access	Get full text
ISSN	1087-0156 1546-1696 1546-1696
DOI	10.1038/s41587-020-0560-x

Cover

More Information
Summary:	The contrast and resolution of images obtained with optical microscopes can be improved by deconvolution and computational fusion of multiple views of the same sample, but these methods are computationally expensive for large datasets. Here we describe theoretical and practical advances in algorithm and software design that result in image processing times that are tenfold to several thousand fold faster than with previous methods. First, we show that an ‘unmatched back projector’ accelerates deconvolution relative to the classic Richardson–Lucy algorithm by at least tenfold. Second, three-dimensional image-based registration with a graphics processing unit enhances processing speed 10- to 100-fold over CPU processing. Third, deep learning can provide further acceleration, particularly for deconvolution with spatially varying point spread functions. We illustrate our methods from the subcellular to millimeter spatial scale on diverse samples, including single cells, embryos and cleared tissue. Finally, we show performance enhancement on recently developed microscopes that have improved spatial resolution, including dual-view cleared-tissue light-sheet microscopes and reflective lattice light-sheet microscopes. Microscopy datasets are processed orders-of-magnitude faster with improved algorithms and deep learning.
Bibliography:	ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 14 content type line 23 Equal contribution Conceived project: M.G., Y.L., H.L., Y.W., H.S. Designed experiments: M.G., Y.L., Y.S., T.L., D.D.N., M.W.M., L.H.D., I.R-S., D.G., A.B., J.C., H.V., V.P., D.C-R., Y.W., H.S. Performed experiments: M.G., Y.L., Y.S., T.L., D.D.N., M.W.M., L.H.D., I.R-S., D.G., A.B., J.C., H.V., S.G., T.B.U., Y.W. Prepared samples: Y.S., T.L., D.D.N., M.W.M., L.H.D., R.I., I.R.-S., A.B., J.C., H.V., T.B.U., Y.W. Built instrumentation: T.L., H.V., Y.W. Developed and tested deep learning algorithms/software: Y.L., H.L., Y.W. Developed new registration, deconvolution algorithms/software: M.G., Y.L., P.J.L., Y.W. Recognized link between medical imaging algorithms and improved deconvolution: P.J.L. Tested new registration, deconvolution algorithms/software: M.G., W.A.M., Y.W. Developed and tested big data pipeline: M.G., Y.S., Y.W. Contributed lineaging/segmentation software and expertise: D.D.N., A.S., Z.B. Contributed samples: C.M.A., M.H., A.B.C. Wrote manuscript: M.G., Y.L., Y.S., P.J.L., Y.W., H.S. with input from all authors. All authors inspected data and contributed to the drafting of the manuscript. Supervised research: V.P., J.C.W., C.M.A., M.H., W.A.M., A.B.C., A.U., T.B.U., Z.B., D.C-R. P.J.L., H.L., Y.W., H.S. Directed research: H.S. Author Contributions
ISSN:	1087-0156 1546-1696 1546-1696
DOI:	10.1038/s41587-020-0560-x